Interpretable Classification And Regression For Time Series Data

Deep architectures have drawn the attention of the researchers working in various areas such as machine learning and computer vision after they have been proposed to the literature. The main reason for this interest is due to their complex mathematical structures and the superior learning abilities resulting from that. What they gain in performance, they lose in interpretability. Extreme Gradient Boosting model offers comparable performance to deep architectures while retaining interpretability. In this paper, this model is applied to time series data by extracting features over those series. The performance of the model is compared with the base models and analyses are made on how the features are used.